HomePhysical DesignDay 1 — Die Planning

Die Planning & Floorplanning

The first step of physical design. Aspect ratio, core utilization, macro placement, power delivery planning, I/O ring design, and production floorplan strategy for VLSI chips.

By EcrioniX Engineering Team · Published June 14, 2026 · ~4,500 words · 14 min read

1. What is Die Planning?

Die planning is the first critical step in physical design. It determines the chip's overall dimensions, aspect ratio, placement zones, and power distribution strategy before any detailed layout begins. Every later step — placement, CTS, routing, timing closure — inherits the consequences of these early decisions.

Key decisions made during die planning:

Why It Matters Most

Poor die planning is the single most expensive mistake in physical design. A weak floorplan can doom timing closure ten steps later — and by then, fixing it means restarting the entire flow. Experienced PD engineers spend weeks on the floorplan because it constrains everything downstream.

2. Aspect Ratio Selection

Aspect Ratio = Die Width / Die Height. It affects routing congestion, macro fit, and I/O perimeter cost. The right choice depends on your macro shapes, package, and dominant routing direction.

Common aspect ratios (100 mm² die): 1.0 (square): 10.0mm × 10.0mm perimeter = 40.0mm → minimal I/O cost 1.5 (wide): 12.2mm × 8.2mm perimeter = 40.8mm → easier H-routing 2.0 (wider): 14.1mm × 7.1mm perimeter = 42.4mm → high congestion 0.67 (tall): 8.2mm × 12.2mm perimeter = 40.8mm → V-friendly, hard route Trade-off: square minimizes I/O perimeter (cost), but rectangular often fits rectangular macros (SRAM) better.

Visual comparison of die aspect ratios:

Aspect Ratio Comparison — Same 100 mm² Area
10×10 Square 1.0 Perimeter 40.0mm Balanced routing 12.2×8.2 Wide 1.5 Perimeter 40.8mm Easier H-routing 8.2×12.2 Tall 0.67 Perimeter 40.8mm Harder H-routing

3. Core Utilization Target

Core utilization = (standard cell area + macro area) / available core area. It's the single most important floorplan knob — too low wastes silicon, too high makes timing closure impossible.

Technology NodeTarget UtilizationRouting CongestionTiming Risk
28nm+60–65%LowLow
14nm70–75%ModerateModerate
7nm75–80%HighHigh
5nm80–85%Very HighVery High

Why not 100% utilization?

4. Macro Placement Strategy

Macros are pre-designed blocks: memories (SRAM), hardmacros (PLL, analog), and embedded processors. Their placement is locked early because they're large and immovable once the floorplan is frozen.

Floorplan with Macro Placement (Top-Down)
I/O Ring (Pads) SRAM 64KB L2 cache bank SRAM 64KB L2 cache bank Core Logic — Standard Cells Utilization 75% PLL Clock gen (analog) PHY I/O interface
Placement principles: co-locate SRAMs near the logic that uses them (minimize wire delay), group memories hierarchically (L1 near CPU, L2/L3 toward periphery), place power-hungry macros symmetrically for even heat spread, and route data buses directly from macro to core.

5. Power Delivery Network (PDN) Planning

The PDN delivers power from package pads to every transistor. A poorly planned PDN causes voltage droop (IR drop), timing violations, functional failures, and electromigration. It's planned hierarchically from coarse (package) to fine (cell).

PDN Hierarchy — VDD Distribution
Package
Bulk capacitors on PCB (100µF) — low-frequency response, 1–2 mΩ
↓ BGA balls
Die (on-chip)
Decoupling caps (1–10µF) — mid/high-frequency, 2–5 mΩ
↓ Power straps (M9/M8)
Cell (local)
Local power rails per cell — ultra-high frequency, 1–3 mΩ
PDN Impedance Budget: Max allowable voltage drop = 10% of supply For a 1.8V supply → max 180mV drop Impedance targets: Package: 1–2 mΩ (bulk power) Die: 2–5 mΩ (distribution) Cell: 1–3 mΩ (local) Total: < 10 mΩ (ensures < 180mV @ 18A peak) Design rule: power straps every 100–200µm, ground return in parallel to minimize loop inductance.

6. I/O Ring & Pad Placement

The I/O ring is the physical boundary holding all pads — power, ground, data, and control. Pad planning balances signal count against power/ground integrity.

Pad Type (256-pin example)CountPurpose
Power (VDD)50Supply current delivery
Ground (GND)50Return path, noise reference
Signal (I/O)100Data, address, control
Spare56Future use / redundancy

7. Core Region Allocation

The core region is the area available for standard cells. A realistic 100 mm² die allocates area across many competing needs:

Typical area allocation (100 mm² die): Macros (SRAM, hardmacro): 20–30 mm² Core logic: 50–70 mm² Routing overhead: 5–10 mm² (metal resources) Power/ground straps: 5–10 mm² (metal area) I/O ring: 5–10 mm²

8. Real-World Floorplan Examples

Mobile Processor (Apple A17)

Server Processor (AMD EPYC)

Die Planning Checklist

Next — Day 2: Power Delivery Network detailed design — IR drop analysis, decoupling capacitors, power strap design, and electromigration.

Course Home
All 15 Days
Next →
Day 2: Power Delivery Networks